Hydraulic performance of biofilter systems for stormwater management

Hydraulic performance of biofilter
systems for stormwater management:
lessons from a field study
Sébastien Le Coustumer, Tim D. Fletcher
Ana Deletic & Matthew Potter
Facility for Advancing Water Biofiltration, Department of Civil Engineering, Institute
for Sustainable Water Resources, Monash University, Melbourne, Vic., 3800, Australia
Supported by the Better Bays and Waterways Institutionalising Water Sensitive Urban
Design and Best Management Practice in Greater Melbourne Project, Melbourne
Water, East Melbourne, 3002, Australia.

This study was undertaken under the Facility for Advancing Water Biofiltration, as part
of PROJECT 4: DEMONSTRATION AND TESTING (ACTIVITY 4.03: Investigation
into the long term sustainability of stormwater bioretention systems). It was supported
by funding through the Better Bays and Waterways - Institutionalising Water Sensitive
Urban Design and Best Practice Management in greater Melbourne project. Funding
partners are the Australian Government through the Coastal Catchment Initiative funded
through the Natural Heritage Trust, Department of Sustainability and Environment,
Melbourne Water, Clearwater and the City of Kingston.
April 2008
Contact details:
Sebastien.le.coustumer@eng.monash.edu.au
Tim.fletcher@eng.monash.edu.au
Ana.deletic@eng.monash.edu.au
matthew.potter@melbournewater.com.au
2

Executive summary
Biofiltration systems (‘biofilters’) are increasingly being used to manage
polluted stormwater runoff in urban areas. However, there are significant concerns
about their lifespan, particularly due to the possibility of clogging of the systems over
time. A study of 37 biofilters constructed on the east coast of Australia during the last
seven years, shows that 60% of constructed systems have a saturated hydraulic
conductivity (K) which meets or exceeds the currently recommended range of 50 to 200
mm/h. It appears that the media used varied greatly between systems, perhaps because
of a lack of available guidelines at the time of construction, or because of inadequate
specification and quality control. However, this generally does not affect the treatment
efficiency of the systems, as most systems surveyed were sufficiently sized (in filter
area or ponding volume) such that their detention storage volume compensates for
reduced media hydraulic conductivity (Figure 1). Consideration of the interaction
between these three design elements – hydraulic conductivity, filter area and detention
(ponding depth) – is critical (rather than consideration of one factor in isolation).
filter media
hydraulic
conductivity
conductivity
extended
detention
depth
infiltration
capacity
filter
surface
area
Figure 1. Interaction between media hydraulic conductivity and other design components, in
determining infiltration capacity of the bioretention system.
The study broadly reveals two types of systems: some with a high initial K
(>200 mm/h) and some with a low initial K (

change occurs over time. Two hypotheses could explain this phenomenon: on one hand
sediment depositions could be leading to the clogging of the surface of the system;
another possibility is that the creation of macropores through root growth and dieback
may help to minimise the reduction in K. The impact of surface clogging is
proportionally greater in systems which started with a high initial K, most likely
because the difference in particle size distribution between the original filter media and
deposited sediments will be greater where the original media was coarse. In the systems
with low initial K, the finer particle size distribution will be more similar to that of the
inflow sediments (although still considerably larger), thus reducing the proportional
impact of any surface clogging effect.
Site characteristics such as filter area as a proportion of catchment area, age of
the system and inflow volume were not found to be useful predictors of media
conductivity, with initial conductivity of the original media explaining the vast majority
of variance. It is clear therefore, that strict attention must be paid to the specification of
original filter media, to ensure that it satisfies current design requirements. Media
should be tested after construction of the system.
Given the apparent difficult in specifying and maintaining hydraulic conductivity
in biofiltration media, one approach is to use a “contingency factor” in the specification
of hydraulic conductivity for biofiltration systems. For example, where the design
intent is to use a soil media with a hydraulic conductivity of 180 mm/hr, sizing of the
system should be undertaken assuming a hydraulic conductivity of 50% of the design
value (ie. 90 mm/hr). In this way, if the media does not meet specifications, or shows a
decline in hydraulic conductivity over time, the overall system performance will remain
satisfactory.
Key words: Biofilters, clogging, stormwater management, hydraulic conductivity
4

Table of contents
1 Introduction .............................................................................................................. 6
2 Method...................................................................................................................... 9
2.1 Sampling locations ........................................................................................... 9
2.2 Sampling methodology..................................................................................... 9
Single ring field infiltration test (shallow test)....................................................... 10
Deep ring field infiltration test ................................................................................ 11
Laboratory infiltration test...................................................................................... 12
Limitations of the sampling methods...................................................................... 12
2.3 Data analysis................................................................................................... 13
3 Results and discussion............................................................................................ 15
3.1 Biofilter design ............................................................................................... 15
3.2 Comparing field methods ............................................................................... 18
3.3 Comparing field and laboratory methods ....................................................... 19
3.4 Clogging of systems over time ....................................................................... 20
3.5 Influence of system characteristics and hydraulic performance..................... 23
4 Conclusions ............................................................................................................ 25
5 Acknowledgments .................................................................................................. 26
6 References .............................................................................................................. 27
5

1 Introduction
Biofiltration systems (also called bioretention systems, biofilters or rain gardens)
have been widely implemented in the last few years as a source control technique to
manage stormwater runoff in urban areas (Melbourne Water, 2005). They form one
technique in the suite of tools which aim to reduce the impact of urbanisation on
watersheds (Dietz and Clausen, 2007).
In Australia, they are commonly used to treat runoff in urbanizing areas and as a
retrofit technique in already-developed areas. With increasing societal pressure to
protect and restore the environmental quality of aquatic ecosystem (Brown and Clarke,
2007), local water authorities are encouraging their development and construction.
Biofilters have been shown to be effective in the treatment of suspended
sediments, heavy metals and nutrients (Zinger et al. 2007). However, there remain
significant questions about the sustainability and long-term performance of biofilters,
with little field data available.
Australian guidelines for biofilter design generally recommend a saturated
hydraulic conductivity (K) of between 50 and 200 mm/h for the filter media (e.g.
Melbourne Water, 2005). Biofilters are then designed in order to achieve pollutant
removal efficiency for total suspended solids (TSS), total nitrogen (TN) and total
phosphorus (TP), as described in Table 1. Curves giving the percentage of pollutant
removal versus the size of biofilter for different detention depth have been developed, to
determine required area and detention volume (Melbourne Water, 2005). As an
example, in the Melbourne area, if a 300 mm detention depth is assumed, a biofilter has
to be sized at approximately 1.65% of the impervious catchment in order to achieve a 45
% TN removal (the limiting performance target required by regulations in several
Australian states).
Table 1: Quality objectives for stormwater treatment in Victoria (Victoria stormwater committee, 1999)
Pollutants
TSS
TP
TN
Objective
80% retention of annual loading
45% retention of annual loading
45% retention of annual loading
6

The potential for clogging of the surface of the systems is a real issue (Bouwer,
2002). A field survey of a number of infiltration systems (which unfortunately did not
include raingardens or any other form of biofilters) conducted by Lindsey et al. (1992)
showed that only 38% of infiltration basins were functioning as designed after 4 years
of operation, with 31% considered to be clogged. Schueler et al. (1992) showed that
50% of infiltration systems were not working due to clogging. Clogging results in more
frequent overflows and therefore a decrease in treatment capacity, since less of the
annual flow volume filters through the media. Clogging may also cause problems of
stagnant water and aesthetic problems, leading to difficulties in acceptance of
biofiltration as a system integrated into the built environment.
To illustrate the impact of the variation of K on the performance of the system, the
Model for Urban Stormwater Improvement Conceptualisation (MUSIC v3.02) software
was used to model a hypothetical catchment of 1 ha (100% impervious) located in
Melbourne, draining into a biofilter designed at 1% of the catchment area, and with a
ponding depth of 10 cm. As shown in Figure 1, the percentage of mean annual flow
treated (called ‘hydrologic effectiveness’; Wong et al. 1999) decreases with K. For a K
of 200 mm/h, the hydrologic effectiveness is 85%; falling to 58% for K of 50 mm/h and
to 29% for K of 5 mm/h. When biofilters are severely clogged (for example when K is 5
mm/h), 71% of inflows are discharged effectively untreated to receiving waters.
Removal of TN load / Hydrological Effectiveness [% of
annual inflows]
100
90
80
70
60
50
40
30
20
10
0
Hydrological effectiveness
TN reduction
1 10 100 1000
Hydraulic conductivity (mm/h)
Figure 1 : Evolution of Hydrological effectiveness and TN reduction vs. Hydraulic conductivity
7

In this context, the objectives of this study are to:
• Evaluate the hydraulic conductivity and the design of a large number of
systems already built in order to assess their performance regarding current
guidelines,
• Assess the evolution of hydraulic conductivity in order to estimate the
variation of their performance with time,
• Gather a large dataset on different systems in order to understand the
possible influence of some factors (i.e. catchment size, biofilter size, age…)
on their performance.
• Provide guidance on the design and construction of rain-gardens, based on
findings of the survey.
8

2 Method
2.1 Sampling locations
Thirty-seven biofilters were sampled over 18 sites, in Melbourne, Sydney and
Brisbane. Measurements were taken in order to represent as well as possible any spatial
variability in hydraulic conductivity. Usually three measurements were taken in each
biofilter, most of the biofilters tested being less than 40 m² in area (Table 2).
The sites were deliberately chosen to have different characteristics. Impervious
catchment and biofilter size were measured for each. The distributions of the main
characteristics of the sites (catchment location, type and size, and biofilter size) are
presented in Table 2. The volume of water, the volume of water received by the system
every year and the total volume of water per m² of biofilter were calculated with the
following hypotheses: an average annual precipitation (Bureau of Meteorology):
• Melbourne: 650 mm/h
• Brisbane: 1200 mm/h
• Sydney: 1175 mm/h
• An annual runoff coefficient of 0.8 (to account for initial losses, etc).
Table 2: Site characteristics
Location # of sites Catchment use # of sites Catchment size # of sites Biofilter size # of sites
Melbourne 30 Residential 18 < 100m² 6 < 40m² 31
Brisbane 4 Industrial 8 100–1000 m² 17 40–400 m² 4
Sydney 3 Parking 5 1000-10000 m² 9 > 400 m² 2
Highway 4 > 10 000 m² 1
Low traffic road 2
2.2 Sampling methodology
A number of alternative methods were used (for comparative purposes) for
determination of hydraulic conductivity. Field hydraulic conductivity (K fs )
measurement of biofilters, have been conducted by two different methods (used for
cross-checking the measurements):
9

(1) a single ring shallow infiltrometer (referred to herein as K fs shallow ), and
(2) a deep ring infiltrometer (herein referred to as K fs deep ).
Laboratory measurements of hydraulic conductivity were also conducted on two
types of samples:
(1) surface samples (K lab surf ), and
(2) samples taken from deep in the filter media (at a depth of approximately 150
mm below the surface - referred to as K lab deep ini )
Whilst the surface measurement reflects the current effective hydraulic
conductivity of the system, the deep sample provides an estimate of the initial hydraulic
conductivity of the media. Methods used for determining each of these measures are
described in more detail in subsequent sections.
Single ring field infiltration test (shallow test)
The single ring infiltrometer test has been widely described in the literature (see
Reynolds and Elrick, 1990, Youngs et al. 1993 for example) and thus only a brief
summary is given here. This method measures the hydraulic conductivity at the surface
of the soil (and thus is most appropriate when the hydraulic conductivity is controlled
by a limiting layer at the media surface).
The single ring infiltrometer consists of a small plastic ring, with a diameter of
100 mm that is driven 50 mm into the soil (Figure 2). It is a constant head test that is
conducted for two different pressure heads (50 mm and 150 mm). The experiment is
stopped when the infiltration rate is considered steady (i.e. when the volume poured per
time interval remains constant for at least 20 minutes).
In order to calculate K fs , a ‘Gardner’s’ behaviour for the soil was assumed
(Gardner, 1958 in Youngs et al. 1993):
K(
h)
= K fs
e
αh
Eq. 1
Where K - the hydraulic conductivity, h - the negative pressure head and α - a soil pore
structure parameter.
10

K fs is then found using the following analytical expression (for a steady flow)
(Reynolds and Elrick, 1990):
H φm
q = K
fs
( 1 + ) +
Eq. 2
πaG
πaG
Where q - the steady infiltration velocity, a -the ring radius, H - the ponding depth, φm -
the matrix flux potential and G a shape factor estimated as:
d
G = 0 .316 + 0.184
a
Eq. 3
Where d - the depth of insertion of the ring, and a - the ring radius. G is considered to be
independent of soil hydraulic conductivity (i.e. K fs and α) and ponding, if the ponding
depth is greater than 50 mm.
Deep test
Shallow test
Constant ponding depth (H)
Possible clogged layer
Media filter (L)
Transition and
drainage layer
To the drainage
system
Figure 2: Sketch of a biofilter with the deep ring and the shallow ring
Deep ring field infiltration test
The deep ring method is based on the lysimetric method as explained in Daniel
(1989), and may be more appropriate if the limiting hydraulic conductivity is found
considerably below the surface. In order to measure the K fs with this method it is
necessary to have a free-draining outflow drain from which the discharge can be
measured. Biofilters are built effectively as a lysimeter (i.e. with a drainage layer made
of gravel and perforated pipe, both of which have a flow capacity several orders of
magnitude higher than the overlying filter media), but it was not feasible to flood each
biofilter and measure the outflow in the drain. However, by inserting a ring down to the
drainage layer (Figure 2), it is then possible to apply Darcy’s law, assuming the
following hypotheses: flow through the soil is uni-directional (vertical); soil is
11

saturated, we can then assume that the wetting front is going down the drainage layer
and will be as thick as the soil layer and there will be no pressure in the drainage layer
(ie. free drainage is assumed).
Darcy’s law (Eq. 4) can be then used to calculate the hydraulic conductivity.
QL
K fs =
( H LA )
+
Eq. 4
Where L - the length of the media filter, H - the ponding depth, Q - the infiltration flow
rate (i.e., added volume of water to keep a constant head divided by the time step) and A
the cross sectional area of the column.
The cylinder used had a 130 mm internal diameter and was driven up to the gravel
layer. This method is able to measure the hydraulic conductivity of the whole system,
and can thus account for a limiting layer anywhere in the media depth profile.
Laboratory infiltration test
Soil samples were tested according to the Australian standard AS 4419-2003.
Tests were undertaken on disturbed samples (by necessity). Surface samples (called labsurface)
were taken in the first centimetre of the soil and deep samples between 10 and
15 cm. Surface samples were used to compare the field and the laboratory method. Deep
samples (called lab-deep) are assumed to represent ‘the initial hydraulic conductivity’
of the systems because soil has not been changed by sediment deposit (any fine
sediment deposited on the filter is assumed to be trapped within the first few
centimetres of the media (Hatt et al, 2006). The dimensions of samples tested during the
experiments were 100 mm diameter and 85 mm deep.
Limitations of the sampling methods
For both field tests, the main limitations are the possible compaction or
disturbance of the soil while the ring is driven into the soil and possible preferential
flow on the side of the column. For the deep ring method, it is also possible that the
assumptions of saturated conditions at the beginning of the experiment, and the freedraining
behaviour of the underlying gravel, may not be true in all cases.
12

Laboratory measurements have inherent problem that samples are in all cases
disturbed, and therefore impact of in-situ soil compaction can not be evaluated.
Although it is clamed that these measurements will give ‘the initial hydraulic
conductivity’, this is not exactly true since soil structure may have been changed over
time (e.g. water flowing through the systems over time should have changed soil
structure to some extent).
2.3 Data analysis
Data preparation
For each biofilter and when multiple tests were conducted, average hydraulic
conductivity and its coefficient of variation were calculated for each method.
Uncertainties were calculated using the law of propagation of uncertainty as explained
in NIST, 1994. Data have been logged transformed before statistical analysis in order to
respect the assumptions of normality, which was tested with Kolmogorov-Smirnov test
(K-S test); normality accepted at p>0.05.
Comparison between methods
The results from the two different field methods are compared as well as field
methods and laboratory methods. Paired t-Tests are conducted with results considered
as statistically different when p< 0.05; coefficient of correlation (R²) between methods
is also calculated. The laboratory – field comparisons are only made for the lab-surface
samples because it is assumed that lab-deep samples represent the initial hydraulic
conductivity and not the current value (Hatt et al., 2006).
Determining clogging over time
As the initial hydraulic conductivity is unknown for all the systems tested, the
measurement made on the lab-deep samples is assumed to represent the initial hydraulic
conductivity (K lab deep ini ). This value is then compared with the field measurements (K fs
deep and K fs shallow ) and with laboratory measurement made on the surface samples (K lab
surf).
13

Determining the influence of catchment characteristics
Ascendant Hierarchical Classification (AHC) was used to show group patterns in
the dataset, and thus to tease out the factors which have significant explanatory power
on the variation observed. It was conducted on the following parameters: K fs shallow , K fs
deep, K lab deep ini , age of the system, size ratio, volume of water per year, total volume of
water and total volume of water per area (m²) of systems.
Multiple regressions were used in order to explain the variation of K fs shallow as a
function of explanatory variables (K lab ini , age, size ratio, volume of water per year, total
volume of water per m²) in order to understand possible correlation between the current
K and characteristics of the system and its catchment.
14

3 Results and discussion
3.1 Biofilter design
Most of the systems studied were not designed strictly according to current
guidelines in term of ratio size of the system/size of the catchment (many were
constructed prior to the availability of these current guidelines). For example, the
Victorian (Melbourne Water, 2005) guidelines recommend a biofilter :catchment area
ratio of at least 1.65 %, but the 30 Victorian systems surveyed show a ratio between
0.1% and 21.9%, with an average of 5.3 % and a median of 2.5% (Table 3). Biofilters
are thus generally over sized, although the extended depths are often considerably less
than the recommended range of 100-200mm. However, given their irregular geometry,
exact measurement of the ‘effective’ or ‘average’ ponding depth was not feasible.
Field hydraulic conductivity measurement shows that 39% (using the shallow ring
method) (42% using the deep ring method) of the biofilters have a K fs below 50 mm/h,
44% (27%) between 50 and 200 mm/h and 17% (30%) above 200 mm/h (Table 3). In
summary, most of the systems (~60%) have a hydraulic conductivity either within or
exceeding the currently used guidelines (Melbourne Water, 2005; Wong et al., 2006).
15

Table 3: Data for each site: hydraulic conductivity value (different method), catchment characteristics and
volume of water received
Hydraulic conductivity (mm/h) Sites characteristics Volume of water
Sites
K fs shallow ur(K) % K fs deep ur(K) %
K lab deep
(ini)
K lab surf
Age (year)
Catchment
area (m²)
Biofilter
area (m²)
Ratio (%)
V. water/year
(m3)
Total V. water
(m3)
Total V./ m²
(m)
Streisand Dr, Brisbane 61 29 32 99 399 238 0.5 1105 20 1.8 1060.8 530.4 26.5
Saturn Cr, Brisbane 34 20 38 10 17 9 0.5 675 20 3.0 648.0 324.0 16.2
Donnelly Pl, Brisbane 19 22 63 12 12 0.2 1130 32.2 2.8 1084.8 217.0 6.7
Hoyland Dr, Brisbane 204 16 667 6 65 146 5 17400 860 4.9 16704.0 83520.0 97.1
58 30 68 14 53 0.5 1500 15 1.0 780.0 390.0 26.0
Monash Car park, Clayton
102 22 88 13 117 0.5 1500 15 1.0 780.0 390.0 26.0
45 20 55 15 48 0.5 1500 15 1.0 780.0 390.0 26.0
71 31 406 12 582 31 3 366 14.5 4.0 190.3 571.0 39.4
594 34 444 10 264 3 60 11 18.3 31.2 93.6 8.5
129 34 265 9 462 98 3 560 4.5 0.8 291.2 873.6 194.1
Cremorne St, Richmond
316 34 282 12 747 116 3 622 18 2.9 323.4 970.3 53.9
98 32 100 17 113 28 3 400 10 2.5 208.0 624.0 62.4
119 25 203 9 297 113 3 324 11 3.4 168.5 505.4 45.9
53 26 202 17 151 4 3 84 6 7.1 43.7 131.0 21.8
85 25 140 11 270 102 3 85 10 11.8 44.2 132.6 13.3
49 27 5 50 10 2 68 12 17.6 35.4 70.7 5.9
Aleyne St, Chelsea
35 34 6 39 20 2 112 24.5 21.9 58.2 116.5 4.8
5 29 9 79 21 2 213 17 8.0 110.8 221.5 13.0
19 30 8 50 12 2 163 22 13.5 84.8 169.5 7.7
Point Park, Docklands
Hamilton St, W. Brunswick
139 25 321 7 246 1 410 7 1.7 213.2 213.2 30.5
135 20 77 47 306 1 370 7 1.9 192.4 192.4 27.5
36 14 7 76 5 3 3200 4 0.1 1664.0 4992.0 1248.0
137 34 11 63 11 3 91 1 1.1 47.3 142.0 142.0
13 36 11 47 9 3 120 5 4.2 62.4 187.2 37.4
Avoca Cr, Pascoe Vale
26 34 10 51 11 3 200 4 2.0 104.0 312.0 78.0
44 34 6 59 15 3 310 4 1.3 161.2 483.6 120.9
24 34 23 30 4 3 314 12 3.8 163.3 489.8 40.8
Parker St, Pascoe Vale
19 21 56 13 16 3 157 14 8.9 81.6 244.9 17.5
39 34 1 138 7 3 528 7 1.3 274.6 823.7 117.7
Ceres, West Brunswick 97 19 60 13 2 1257 21.75 1.7 653.6 1307.3 60.1
Bourke St tree pit,
84 23 23 21 1 100 1.44 1.4 52.0 52.0 36.1
Melbourne
Hallam Bypass, Floret Pl 154 24 199 63 3 120
Hallam Bypass, Wanke Rd 115 22 490 387 3 12
Hallam Bypass, Wanke Rd
basin
Wolseley Pd, Vic Park
(NSW)
Leyland Gr, Vic Park
(NSW)
203 27 286 207 3 168
159 18 425 5 376 560 7 1504 330 21.9 1413.3 9893.0 30.0
237 17 398 4 151 224 7 1804 180 10.0 1695.8 11870.3 65.9
2 nd Pond Creek (NSW) 5 21 2000 0
16

The influence of this result on the system hydrologic effectiveness is shown by
the MUSIC model results in Figure 3. For a system designed at 2.5 % of its catchment
area in Melbourne (which is large by current standards), an extended detention depth of
30 cm and with a K of only 50 mm/h, 74% of the mean annual inflow is treated by the
biofilter. If the system conformed with the recommended hydraulic conductivity (K of
180 mm/hr), it would intercept only 18% more of the mean annual flow. In other words,
the observed frequent over-dimensioning of filter area compensates for the low
hydraulic conductivity, even through detention depths are generally relative shallow in
the systems tested. When K fs drops to 5 mm/h, 44% of flow is treated (for a size of
2.5%) and 64% for a size of 5.3%. The key lesson is that over-sizing of biofilter area
will help to ‘buffer’ against unintended reductions in hydraulic conductivity.
Flow treated (%)
100
90
80
70
60
50
40
30
20
10
0
Biofilter of 2.5% of Catchment
Melbourne
Sydney
Brisbane
0 50 100 150 200 250 300 350 400 450 500
Ks
K(mm/h)/ Size ratio (%)* 1.65 2.5 5.3
180 85 92 98
50 63 74 90
10 40 51 71
5 34 44 64
* With Melbourne data
Figure 3: Influence of the ratio biofilter/catchment size and of the K fs on the ‘hydrologic effectiveness’
At the time of this study, no Australian guidelines made explicit
acknowledgement of this issue; the interaction of hydraulic conductivity, filter area and
ponding depth, in determining the effective infiltration capacity of the biofiltration
system (Figure 4). However, in response to this lack, the Facility for Advancing Water
Biofiltration (FAWB) recently revised its filter media guidelines, to provide greater
guidance on the interaction between design elements in determining infiltration capacity
(Figure 4). Designers should not consider hydraulic conductivity in isolation, but
rather, should design systems using an integrated approach.
17

filter media
hydraulic
conductivity
conductivity
extended
detention
depth
infiltration
capacity
filter
surface
area
Figure 4. Design elements that influence infiltration capacity (source: www.monash.edu.au/fawb).
3.2 Comparing field methods
Statistically both methods give similar results (p=0.38 for n=32). Correlation
between two methods is moderately strong with R² = 0.44 (Figure 5). This result can be
explained by the similarity between the two methods, which are both a pressure test
applied on the same surface of soil. The differences could be explained by the
uncertainty on each reading (around 30% as shown in Table 3) and by the spatial
variability in the systems. Contribution to uncertainty of the latter can be evaluated at
50% as shown in Table 4, based on repeated application of the hydraulic conductivity
measurements in one system. Since tests have not been conducted in the same spot it
could explain the moderate correlation.
600
Kfs deep (mm/h)
400
200
y = 1.00x + 40.53
R 2 = 0.44
K fs shallow K fs deep
Mean (mm/h) 100 140
σ (mm/h) 115 172
Cv (%) 115% 123%
n 32
p 0.38*
* logged transformed data to respect the normality for K fs deep
0
0 200 400 600
Kfs shallow (mm/h)
Figure 5 : Correlation between field methods
18

Table 4 : K value and Cv for each site
Shallow method
Deep method
Sites n K (mm/h) Cv (%) n K (mm/h) Cv (%)
Cremorne St, pod7, Richmond 3 119 29 3 203 32
Cremorne St, pod9, Richmond 3 85 55 3 140 54
Parker St, pod 2 Pascoe Vale 3 19 46 3 56 48
Ceres, Brunswick East 4 97 63 4 60 73
Hamilton St pod1, Brunswick West 3 36 52
Monash car park, strip 1 4 58 34
Monash car park, strip 2 3 102 58
Monash car park, strip 3 3 45 33
Point park car park, Docklands, pod2 3 135 28
Alleyene st, Chelsea, pod4 3 19 140
Tree pit, Docklands, pod7 3 84 78
Hallam Bypass, Wanke Rd 3 115 55
Hoyland St 6 204 59 3 667 32
Donnelly Pl 3 19 58
Saturn dr 3 34 16 3 38 67
Wolseley Pd, Vic Park 5 159 63 5 425 80
Leyland Gr, Vic Park 5 237 55 5 398 21
2nd Pond Creek 9 5 67
Average 78 53 221 53
3.3 Comparing field and laboratory methods
The shallow field method gives results that are not statistically different from the
measurement made in the laboratory on samples taken from the surface (p=0.71, n=16).
The consequence of this is that there is very little bias introduced by the choice of
method. It is therefore possible to compare measurement from the field (shallow
method) and from the lab without effect of the method. These results are similar to that
of Reynolds et al.(2000), which showed for a sand and a loam, that the hydraulic
conductivity measured in the field (shallow method) does not vary significantly from
laboratory measurements (p>0.05, n=12 for the sand, n=10 for the loam).
19

The correlation between methods is low (R²=0.08). It can be explained by the
spatial variability of the hydraulic conductivity and by the variations in compaction
between the samples. In the laboratory, measurements are conducted on disturbed
samples (inevitably during sampling) that have been re-compacted to a standard value
which may be different than that which existed in the field. It may also be speculated
that another reason for lower K found in the laboratory measurements is due to the
samples being taken right at the surface; they therefore represent deposition formed over
time (i.e. they are deposited stormwater sediment). In this way K lab, surf represents the
hydraulic conductivity of the clogging layer on its own.
If we compare the results from the deep field tests and the laboratory tests, results
are statistically different (p

(2003). Since both field measurements of the hydraulic conductivity give similar results
(average value: K fs deep = 140 mm/h, Cv=123% and K fs shallow =100 mm/h, Cv=115%), it
is evident that hydraulic conductivity of the system is controlled primarily by the top
layer and that there is no deep ‘clogging’ of the soil media.
However, vegetation development and especially root growth, will lead to the
creation of macropores. For example, Archer et al. (2000) showed that root growth
increases hydraulic conductivity, as root dieback creates macropores which facilitate
water movement in the soil. It is not yet clear whether this phenomenon will have a
major impact on biofilter hydraulic conductivity; if clogging is primarily occurring on
the surface, macropores below the clogged layer at the top may have little or no
consequence.
Results of the Ascendant Hierarchical Classification show four groups with
distinctly different behaviour. Group 1 has only one biofilter, which is undersized (0.1
% of the catchment); group 2 has three systems with very high K fs (K fs shallow average =
200 mm/h) and very high initial K (K lab deep ini average = 197 mm/h). Groups 3 and 4
represent 88% of the systems tested. Biofilters from group 3 have a high initial K (K lab
deep ini average = 241 mm/h, n=17), whilst group 4 systems have a low initial K (K lab deep
ini average= 12 mm/h, n=11).
Systems with a high initial hydraulic conductivity (which can be explained by
media with relatively coarse particles and a subsequently large pore space) will decrease
substantially over time, and proportionally by a greater amount than will systems with a
low initial hydraulic conductivity. This is demonstrated by the fact that the field
shallow test results show a hydraulic conductivity on average 114 mm/h (n=17) lower
than the laboratory deep tests as shown on Figure 7. This result is also confirmed by the
difference between the laboratory tests taken on the deep and surface samples, with an
average difference of 255 mm/h (n=9).
21

Figure 7 : K lab deep ini vs. Kf s shallow – white triangle, system with low initial hydraulic conductivity, black
square, systems with high initial hydraulic conductivity
This decrease can be explained by sediment deposition at the surface. However,
final hydraulic conductivities are still relatively high (K fs shallow = 127 mm/h, n=17), and
likely to be adequate to ensure good pollutant removal performance. This observation
may be either because the systems are only partially clogged, or because creation of
macropores is having some effect in creating flow through the media, possibly even at
the surface (for example, at the base of plant stems, where growing, senescence and
even stem movement due to wind, may cause ‘breaking up’ of any clogging layer)
(Figure 8).
K lab deep ini = 241
K fs shallow = 127 mm/h
Figure 8 : Schematic representation of the evolution of the behaviour of biofilters with high initial K
22

Systems with low initial hydraulic conductivity (explained by a high concentration in
fine particles and thus a low pore volume) show effectively no decrease over time (ΔK
average = +25 mm/h, n=11; Figure 7). In part, this is because the relative difference in
particle size of the filter media, and of the influent sediment, will be less, meaning that
any buildup of sediment at the surface will have proportionally less impact. The slight
increase could again be contributed to by macropore creation by roots (Figure 9),
although further studies are required to test this hypothesis.
K lab deep in = 12 mm/h
K fs shallow = 37 mm/h
Figure 9 : Schematic representation of the evolution of the behaviour of biofilters with low initial K
3.5 Influence of system characteristics and hydraulic performance
Of all the factors tested – age of the biofilter, its initial hydraulic conductivity, the
ratio between its size and the size of the catchment drained, the volume of water
received per year and the volume of water received per m² of system since its
construction – only the initial conductivity provided a statistically significant
explanation of variability in current conductivity of the systems (Table 5).
23

Table 5: Regression between K fs shallow and various parameters
Field method
K fs shallow
R² 0.52
Parameters
P value
(Constant) 0.00
K lab deep (ini)
0.00
Age 0.24
Ratio 0.34
Volume of water/year excluded
Total volume/m² 0.49
Achleitner et al. (2006) reported a similar lack of correlation between hydraulic
conductivity and site characteristics, and made the hypothesis that current K was mainly
governed by the initial value. This result is in some ways unfortunate, because it
provides little guidance to those charged with the maintenance of such systems, in being
able to predict their lifespan and maintenance requirements. However, it does show the
importance of correctly specifying the filter media during the biofilter design, and of
having appropriate quality control to ensure that the supplied and installed media meets
these specifications.
One approach is to use a “contingency factor” in the specification of hydraulic
conductivity for biofiltration systems. For example, where the design intent is to use a
soil media with a hydraulic conductivity of 180 mm/hr, sizing of the system should be
undertaken assuming a hydraulic conductivity of 50% of the design value (ie. 90
mm/hr). In this way, if the media does not meet specifications, or shows a decline in
hydraulic conductivity over time, the overall system performance will remain
satisfactory.
24

4 Conclusions
Whilst biofilters have been demonstrated to provide effective stormwater quality
treatment, their long-term hydraulic behaviour has to date not been studied, particularly
in reference to real systems. This study provides a first attempt to evaluate performance
of a range of constructed systems, with design and catchment characteristics.
From a measurement point of view, the different field methods used gave similar
results, demonstrating that for these soil-based biofilters, hydraulic conductivity is
governed by their surface layer. Field and laboratory experiments gave identical
results.
Regarding system design and construction, three key messages are deduced.
Firstly, whilst many systems measured have a low hydraulic conductivity (lower
than currently recommended values), a tendency by designers to over-dimension the
systems (relative to guidelines) acts to compensate for the low conductivity. Critically,
however, current guidelines do not address this relationship, and seem to pay no
attention to the risks of diminished effectiveness when systems are either constructed
with lower-than-desired conductivity, or when clogging causes conductivity to decline
over time. In particular, this may occur as a result of poor construction management
practices in the catchment, resulting in excessive sediment loading. Strict controls
should be in place during the construction phase of development.
Secondly, proportional hydraulic conductivity reduction occurs mainly for
systems with high initial value, but the resulting value ends up generally respecting the
guidelines. Other systems, which have been constructed with low-conductivity soils, do
not show evidence of further decline, possibly because the filter media particle size
distribution is more similar to that of the influent sediment, than is the case for systems
with high initial conductivity (and thus coarse media). Declines in conductivity over
time are likely to occur by sediment deposition, which occurs at the surface of the
systems. Whilst macropore creation by vegetation may limit the effect of clogging,
further detailed research is needed to verify the reliability of this strategy in maintaining
soil hydraulic conductivity within recommended guidelines.
25

Finally, it was not possible to predict a filter’s current hydraulic conductivity from
factors such as its size, the catchment size, or the inflow volume. The initial specified
hydraulic conductivity is the critical determinant of its long-term hydraulic behaviour.
Whilst this provides little help in predicting system lifespan or maintenance
requirements, it does reinforce the criticality of specifying the correct hydraulic
conductivity of systems at the time of construction. Perhaps most importantly,
guidelines do not pay due attention to the importance of translating design
specifications through the construction process. Contract hold-points should in place to
ensure testing of the media during construction.
5 Acknowledgments
The authors would particularly like to thank Peter Poelsma at the Facility for
Advancing Water Biofiltration for help in the field experiments. Support from
Melbourne Water is also gratefully acknowledged.
26

6 References
Achleitner S., Engelhard C., Stegner U. and Rauch W. Local infiltration devices at parking sites-
Experimental assessment of temporal changes in hydraulic and contaminent removal capacity, WSUD
2006, Melbourne
Archer N. A., Quniton J. N. and Hess T. M. Below ground relationships of soil texture, roots and hydraulic
conductivity in two-phase mosaic vegetation in South-east Spain. Journal of Arid Environments, 2002,
Vol. 52, p 535-553
AS 4419—2003, Australian Standard, Soils for landscaping and garden use, appendix H, Method of
determining the permeability of a soil.
Bouwer Herman. Artificial recharge of groundwater: hydrogeology and engineering. Hydrogeology Journal,
2002, Vol. 10, p 121-142.
Brown R. and Clarke J. The transition towards water sensitive urban design : a socio-technical analysis of
Melbourne, Australia, Novatech 2007 Sustainable techniques and strategies in urban water
management, Lyon, France, June 2007, p 349- 356
Bureau of Meteorology, < www.bom.gov.au>
Daniel D. E. In situ hydraulic conductivity tests for compacted clay. Journal of Geotechnical Engineering,
1989, Vol. 115 (9)
Dietz M.E. and Clausen J.C. Stormwater runoff and export changes with development in a traditional and
low impact subdivision. Journal of Environmental Management (2007) (sumitted)
Gardner W.R. Some steady-state solutions of the unsaturated moisture flow equation with application to
evaporation from a water table. Soil Sci., 1958, Vol. 85, p 228–232.
Hatt, B. E., N. Siriwardene, A. Deletic and T. D. Fletcher. Filter media for stormwater treatment and
recycling: the influence of hydraulic properties of flow on pollutant removal. Water Science and
Technology, 2006, Vol. 54(6-7), p 263-271.
Langergraber G., Haberl R., Laber J. and Pressi A. Evaluation of substrate clogging processes in vertical
flow constructed wetlands. Water Science an Technology, 2003, Vol. 48(5), p 25 -34
Lindsey G, Roberts L. and Page W. Inspection and maintenance of infiltration facilities, J. Soil and Water
Cons., 1992, Vol. 47 (6), p 481-486
Melbourne Water (2005). Water Sensitive Urban Design (WSUD) Engineering procedures: stormwater.
Chapter 6 Bioretention basins CSIRO publishing, 2005, Melbourne
National Institute of Standards and Technology (NIST), United States Department of Commerce
Technology Administration, Technical Note 1297, Guidelines for Evaluating and Expressing the
Uncertainty of NIST Measurement Results (1994)
Reynolds W. D. and Elrick D. E. Ponded infiltration from a single ring: Analysis of steady flow. Soil Sci.
Soc. Am., 1990, Vol. 54, p. 1233-1241
Reynolds W. D., Bowman B. T., Brunke R. R., Drury C. F. and Tan C. S. Comparison of tension
infiltrometer, Pressure infiltrometer, and soil core estimates of saturated hydraulic conductivity. Soil Sci.
Soc. Am., 2000, Vol. 64, p 478-484
Schueler, T.R., Kumble, P.A. and Heraty, M.A. A Current Assessment of Urban Best Management
Practices, Techniques for Reducing Non-point Source Pollution in the Coastal Zone, Metropolitan
Washington Council of Governments, 1992, Washington, D.C.
Victoria stormwater committee. Urban Stormwater Best Practice Environmental Management Guidelines.
CSIRO publishing, 1999
Winter K.-J. and Goetz D. The impact of sewage composition on the soil clogging phenomena of vertical
flow constructed wetlands. Water Science and Technology, 2003, Vol. 48(5), p 9-14
27

Wong, T. H. F. (Ed.). (2006). Australian runoff quality. Sydney, Australia: Institution of Engineers, Australia.
Wong T. H. F., Breen P. F., Somes N. G. L. and Lloyd S. D. Managing Urban Stormwater Using
Constructed Wetlands. Melbourne, Australia: CRC for Catchment Hydrology and CRC for Freshwater
Ecology Industry Report 98/7, 1999
Youngs E. G., Elrick D. E. and Reynolds W. D. Comparison of steady flows from infiltration rings in “Green
and Ampt” and “Gardner” soils. Water Resources Research, 1993, Vol. 29 (6), p 1647-1650
Zinger Y., Fletcher T. D., Deletic A., Blecken G. T. and Viklander M. Optimisation of the nitrogen retention
capacity of stormwater biofiltrations systems, Novatech 2007 Sustainable techniques and strategies in
urban water management, Lyon, France, June 2007, p 893- 900
28